一种改进的卷积神经网络的无参考JPEG2000图像质量评价方法
Blind JPEG2000 Image Quality Assessment via Improved Convolutional Neural Network
卷积神经网络(Convolutional neural network)是当前图像处理领域的研究热点。本文提出了一种基于卷积神经网络的JPEG2000压缩图像质量评价方法。该模型由一层包含20个卷积核的卷积层，一层包含最大池、中值池和最小池的次采样层、一层采用1200个ReLU激活单元的全链接层和一个输出节点构成。采用最大、中值、最小三池联合的方法，可以有效提取图像的质量感知特征。在LIVE图像质量评价库JPEG2000压缩图像上的实验结果表明，我们提出的方法得到了比相关文献方法更好的主观感知一致性。
onvolutional neural network(CNN) currently becomes research focus in image processing field. In this paper, we put forward one kind of JPEG2000 compressd image quality evaluation method based on improved CNN framework. The model is consisted of one convolutional layer with 20 convolution kernels ,one pooling layer with the max,median and min pooling, one fully connected layer with 1200 the ReLU units and one output node.Using the max, medium and min pool structure , to effectively extract the typical features in the image.Experimental results show that the proposed method is consistent with the subjective score better in the LIVE library.
李朝锋、朱睿
计算技术、计算机技术电子技术应用
卷积神经网络深度学习无参考图像质量评价
onvolutional neural network (CNN)eep learningNo reference Image quality assessment
李朝锋,朱睿.一种改进的卷积神经网络的无参考JPEG2000图像质量评价方法[EB/OL].(2015-11-11)[2025-08-02].http://www.paper.edu.cn/releasepaper/content/201511-159.点此复制
评论